Machine Learning Accelerated Simulation of Charge Regulation Effects
Abstract
Many solvated materials, from nanoparticles to polyelectrolytes, acquire charge through the dissociation or association of surface groups, a phenomenon known as charge regulation (CR). Recent methods use hybrid charge regulation Monte Carlo (CR-MC)/molecular dynamics (MD) to evolve the system while dynamically adjusting charges on the surface groups. However, this approach is computationally expensive, as Monte Carlo sampling can be as costly as the accompanying MD. In this work, we introduce charge regulation machine learning (CR-ML), a graph attention network that predicts the ionization probability of every surface site from the instantaneous configuration, the dissociation constant of surface groups, and the solution conditions. We show that CR-ML reproduces the equilibrium conformations and degree of ionization obtained with CR-MC for a hydrophobic polyanion. We further test CR-ML on the more complex system of two nanoparticles mediated by a polyelectrolyte, where it accurately predicts the charge asymmetry that develops between the two particles. Across these systems, CR-ML updates charges an order of magnitude faster than CR-MC while accurately reproducing its equilibrium properties. This framework opens a path toward large-scale simulations of charge-regulating materials for understanding and designing their responsive behavior.
Related
- Passage of particles through matter and the effective straggling-function: High-fidelity accelerated simulation via Physics-Informed Machine Learning
- Machine Learning Accelerated Plasma Simulation through Physics Guided Time Jumps
- Learning to Rank Tensor Network Contraction Plans for GPU-Accelerated Quantum Circuit Simulation
- SCALE: Simulation-Calibrated Amortized Learning for Energy Materials (A hybrid architecture connecting deterministic modeling, real-world data, and transformer-scale inference for accelerated energy-materials discovery)
- Machine-Learning-Accelerated Metallene Stabilization from High-Throughput Sandwich Modeling
BibTeX
@misc{yadav2026machine,
title = {{Machine Learning Accelerated Simulation of Charge Regulation Effects}},
author = {Vijay Yadav and Sunita Kumari and Prabhat K. Jaiswal},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.32713},
eprint = {2610.32713},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.32713},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Machine Learning Accelerated Simulation of Charge Regulation Effects
AU - Vijay Yadav
AU - Sunita Kumari
AU - Prabhat K. Jaiswal
PY - 2026
DA - 2026/10/23
PB - arXiv
DO - 10.48550/arxiv.2610.32713
UR - https://arxiv.org/abs/2610.32713
AB - Many solvated materials, from nanoparticles to polyelectrolytes, acquire charge through the dissociation or association of surface groups, a phenomenon known as charge regulation (CR). Recent methods use hybrid charge regulation Monte Carlo (CR-MC)/molecular dynamics (MD) to evolve the system while dynamically adjusting charges on the surface groups. However, this approach is computationally expensive, as Monte Carlo sampling can be as costly as the accompanying MD. In this work, we introduce charge regulation machine learning (CR-ML), a graph attention network that predicts the ionization probability of every surface site from the instantaneous configuration, the dissociation constant of surface groups, and the solution conditions. We show that CR-ML reproduces the equilibrium conformations and degree of ionization obtained with CR-MC for a hydrophobic polyanion. We further test CR-ML on the more complex system of two nanoparticles mediated by a polyelectrolyte, where it accurately predicts the charge asymmetry that develops between the two particles. Across these systems, CR-ML updates charges an order of magnitude faster than CR-MC while accurately reproducing its equilibrium properties. This framework opens a path toward large-scale simulations of charge-regulating materials for understanding and designing their responsive behavior.
ER -